Ivan,
Thank you for that – it is very helpful.

For those collecting info –
We are running our Guacamole system on a two server
VMware ESXi, 6.7.0 setup.
Server 1 is an HP ProLiant DL380 G7
with Dual Intel(R) Xeon(R) CPU E5620 @ 2.40GHz
and 140 Gigs of Ram - which hosts 47 Virtual systems

1 - production web server
1- Guacamole Server - running on a CentOs7
Set with
8 virtual CPU
16 gig Memory
a 100 gig Hard disk
and 1 virtual Nic

40 virtual windows 7 desktops being connected through Guac RDP

Server 2 is an HP ProLiant DL360 Gen10
With Dual Intel(R) Xeon(R) Gold 5218 CPU @ 2.30GHz
and 580 gigs of Ram currently hosting 65 Virtual systems

1 Development Sql server and
64 virtual windows 7 desktops being connected through Guac RDP

We have also added 75 new Guacamole RDP to Physical desktops (mostly Windows 
10) since Work From Home began.
ESXI Hosts are using about 20% of theis CPU resources
We are sitting at about 50% host memory usage
My Guacamole system averages about 30% memory and cpu usage to allocation
We have seen a few more of our international users complain about system speed 
and Guacamole connection instability since moving from our offices to their 
homes, but I feel most of that is due to their home internet connections being 
slower than their office ones were.

We have also been asked to start adding between 50 and 100 new Virtual Windows 
7 users to our system over the next three months. We are confident that we can 
manage this primarily on Server #2

From: ivanmarcus <[email protected]>
Sent: Sunday, April 5, 2020 7:05 PM
To: [email protected]; Jason Keltz <[email protected]>
Subject: Re: hardware requirements for Guacamole


I mentioned before that I'd see if I could separate the Guacmaole traffic from 
the rest of the network, and post the results.

Here's a 1-hr plot of Guacamole only traffic for ~85 users at a reasonably busy 
part of the day. Also included a snapshot showing some detail on the server 
load, CPU & memory usage etc.

What you see is fairly typical, as previously described, as I type this the 1-h 
server load average is somewhat less (1.8), instantaneous load is 1.5 and 1h 
average traffic is also down - but it could be the other way at any one time.

FWIW this is a VM running on an i7 host with 16 cores/64GB of which 6 
cores/16GB are allocated to this VM. Ubuntu 18.04 on host and guest, host has a 
number of other VM's running under Vbox. I deliberately over-spec'd the 
Guacamole VM resource because it's presently mission critical to this client 
and I didn't want anything to get in its way.

It may also be worth commenting that while some people might cringe at using 
Vbox I've found it to be very reliable and used it for this situation because 
others may need to do something on it occasionally and it's user interface is 
fairly simple. Uptime for the previous Guacamole VM (and host) was approx 
18months, this system was new, in reaction to the rapidly escalating need to 
have the entire company working remotely, hence the low current uptime.

Thus from my experience I'd suggest that Mike's simple rule of thumb seems 
pretty good, and that for 300 users if I were to allocate 12 cores/24GB I'd 
expect it to work reasonably well. At that number of users I might consider a 
standalone box with similar specs to my present host (perhaps just 32GB RAM) 
which would nicely give it some extra headroom.

On 5/04/2020 12:58 p.m., Jason Keltz wrote:

Thanks Chris, and Ivan,

So far, it seems that the load requirements are surprisingly fairly minimal.  
The type of standalone server that I'm looking at getting for this task would 
be more than enough to handle this load.

That being said, if anyone else has specs on their systems with even larger 
numbers of users, please keep them coming.  It's very interesting to me.

Jason.

On 4/4/2020 6:43 PM, ivanmarcus wrote:

I expect like many of us who administer systems I've gone from having a handful 
of remote users (across several different companies) to a much larger number in 
a matter of days. Just to add to the general body of knowledge I cite one 
example here 'cos I've kept a particular eye on it.

With approx 80-85 users, of which circa 35 or a little more are doing intensive 
CAD type work and the rest a mix of documents incl PDF images, the average 
bandwidth when busy has been less than 20Mb/s, server load is typically in the 
range 1.5-2 (6 cores allocated) and memory use up to 8GB (incl OS) out of 12GB 
allocated. Very rarely server load will get a little beyond 3, and bandwith 
will burst to 100Mb/s.

I'm not collecting stats directly from the various Guacamole VM's I've got 
running (I guess I could, but am reluctant to mess with them much at this 
time), but attach a 1-day plot of network traffic from one site here. Bear in 
mind this will also include outgoing traffic from the various machines, and a 
busy mailserver, not just Guacamole <-> remote.

I'm not sure but I may be able to separate the Guacamole server traffic out 
from the rest and plot that, will look at it next week and if so I'll provide 
that detail FYI.

On 5/04/2020 3:27 a.m., Chris Misztur wrote:
Hey Jason, yeah compute requirements are fairly low.  As far as network,  20 
RDP users for me have been under 1Mbit outbound total.  There are bursts.  if 
user starts watching YouTube on their remote session then it jumps to 7Mbits 
for that session.

Had same experience running on a RaspberryPi 4.  However, lag was present.  
Compute and network was fine so I’m not sure why.  Video core?

Chris





On Apr 4, 2020, at 9:50 AM, Jason Keltz 
<[email protected]><mailto:[email protected]> wrote:
Hi.

The other day, I asked a question about load balancing on Guacamole.  Nick 
provided useful information in that respect. Thanks again Nick!    On the other 
hand, I'd like to get some idea of how many desktop connections a single 
Guacamole server could handle.  There's really very little information about 
that online (that I can find).  I understand that different users will all do 
different activities (text editing versus YouTube steaming), and that's going 
to have an impact on the numbers, but really, I'm just looking for averages. I 
saw one single mention online to this: 1 core and 2G of memory for 25 users.  
If that's correct, that's really amazing!   Although, I didn't see any mention 
of the network requirements.  I'd like to build a server capable of handling 
150-300 simultaneous desktop connections.  If I could do it with one server, it 
would be great.  Sure, I can load it with dual Xeon processors, and gigabytes 
of memory, but how much network do I need? Is 1 x 10G enough? and will it 
really be able to handle the average load?

Thanks for any feedback you can provide.  You'll be helping me, but you'll also 
be helping others who are trying to search online for the limitations of 
Guacamole.

Jason.

---------------------------------------------------------------------
To unsubscribe, e-mail: 
[email protected]<mailto:[email protected]>
For additional commands, e-mail: 
[email protected]<mailto:[email protected]>



________________________________

The information contained in this message is intended only for the recipient, 
and may be a confidential attorney-client communication or may otherwise be 
privileged and confidential and protected from disclosure. If the reader of 
this message is not the intended recipient, or an employee or agent responsible 
for delivering this message to the intended recipient, please be aware that any 
dissemination or copying of this communication is strictly prohibited. If you 
have received this communication in error, please immediately notify us by 
replying to the message and deleting it from your computer. S&P Global Inc. 
reserves the right, subject to applicable local law, to monitor, review and 
process the content of any electronic message or information sent to or from 
S&P Global Inc. e-mail addresses without informing the sender or recipient of 
the message. By sending electronic message or information to S&P Global Inc. 
e-mail addresses you, as the sender, are consenting to S&P Global Inc. 
processing any of your personal data therein.

Reply via email to